Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
OpenAI Alleges DeepSeek Used Its Models for AI Training
- Thread starter MacRumors
- Start date
- Sort by reaction score
Nope, all GPTs including OpenAI are based on transformer architecture invented by Google. GPT2 is a 1.8 billion parameter model trained on 8 million web pages. It is laughably bad compared to what models are available now. I can run a 4 billion parameter model on my iPhone 13 pro a few years ago.
Deepseek V3 is a 671 billion parameter MoE model, they are not even in the same league. Their CoT model is a distillation from the 671 billion parameter model.
Is there much point in debating about AI theft?
Industry wont respect IP.
China won't prosecute.
Seemingly, neither will the US.
Even if they did, with models made public - can't exactly wind back the clock.
It's a lawless, cut-throat industry but.. what are they gonna do, cry harder?
Industry wont respect IP.
China won't prosecute.
Seemingly, neither will the US.
Even if they did, with models made public - can't exactly wind back the clock.
It's a lawless, cut-throat industry but.. what are they gonna do, cry harder?
I know you’re (presumably?) making a joke, but they actually do the same thing on their own models. Obviously it’s not against the terms of service when they do it, because it’s their model, but they definitely do it.
Did you expect ethic from a Chinese company ? really ? It is their business model, so it is quite normal they did so. But it worked…
And by the way we should ask how Google and OpenAI trained their model since the beginning … with OUR data.
And by the way we should ask how Google and OpenAI trained their model since the beginning … with OUR data.
A quick google search brings up paper, gunpowder, printing, the compass, alcohol, Go and Mahjong, playing cards…..
It’s China so they’re going to steal it. That being said, it’s literally called OpenAI. And now they going “for profit” after virtue signaling that they wouldn’t. They can sit this one out, Sam Altman especially.
One, that is not correct. Just because I didn’t lock my house doesn’t give you the right to go in and take anything you want. There are terms of service. Two, OpenAI trained their model on data that was behind passwords and paywalls, so even that didn’t help.
People’s idea of “ethic” varies. I do not expect “ethic” from any global corporation at this point, least of all OpenAI.
https://www.livescience.com/planet-...ating-steady-loop-of-plasma-for-1-000-seconds Put that in your pipe and smoke it.
No, it's not, unless you know the address directly. There are/were no outside links, but crawlers were still finding them.
What I'm complaining about, is that we didn't give permission for it to be used. We didn't want it scraped by search engines, and don't want it added to AI training sets. Since these AI companies think our opinions matter less than their profit, we've completely removed from the internet.
My wife is an attorney and she doesn't care either.
This was proprietary information, collected by us over 40+ years. We didn't want it index, scraped, used or copied outside of the caving community. The reasons aren't important, though they are many, but the point is that we should be able to choose for it to not be used, or not used, and said companies should respect that.
Last edited:
How ironic that an AI company caught stealing data from other companies to train it's AI is now complaining that a rival AI company stole it's data so it could train the rival companies AI.
It so laughable.
It so laughable.
you used DeepSeek for a few hours!!!! Chine CCP now has everything about you, personal and none personal.
Oh cry me a river, does anyone have sympathy?
Samsung tears into iPhones, Apple does the same to Galaxy phones...
Heck- food scientists and culinary researchers reverse engineer recipes all the time just to recreate them.
That’s how the world moves forward and improves.
Jog on, Altman.
Samsung tears into iPhones, Apple does the same to Galaxy phones...
Heck- food scientists and culinary researchers reverse engineer recipes all the time just to recreate them.
That’s how the world moves forward and improves.
Jog on, Altman.
OpenAI is just a name. There isn’t anything open in their offer.
Fair enough
By the way, for those that get riled up because Deepseek doesn’t want to answer critical questions about Chinese politics, here’s what I got from Google Gemini when asking it whether Trump was a good president:
“I can't help with responses on elections and political figures right now. I'm trained to be as accurate as possible but I can make mistakes sometimes. While I work on improving how I can discuss elections and politics, you can try Google Search.”
Same answer for Biden, Clinton and Bush. But, it was perfectly happy to give an elaborate answer on Reagan and JFK. So if you want to claim Deepseek is Chinese censorship, Google is American censorship.
“I can't help with responses on elections and political figures right now. I'm trained to be as accurate as possible but I can make mistakes sometimes. While I work on improving how I can discuss elections and politics, you can try Google Search.”
Same answer for Biden, Clinton and Bush. But, it was perfectly happy to give an elaborate answer on Reagan and JFK. So if you want to claim Deepseek is Chinese censorship, Google is American censorship.
I never really saw the point of asking an AI for an opinion. I can see asking for something like a background briefing (and even that can be biased) but not whether someone is good or bad. We can't let these things think for us.
Saying this is crazy when OpenAI’s whole business model relies on plagiarism and training off of online content without consent.
